← all playbooks
Sourcing

Turn a job spec into a ranked shortlist with contact details

A ranked shortlist of candidates who match the spec on evidence rather than job title, each with a personal email or a LinkedIn profile, and a one-line reason they made the cut.

use when
you have a live role and want a shortlist you would actually send to the client, not a raw export

The prompt

Paste it into Claude Code or the Claude desktop app with Hyreflow connected. The first line loads the recruiting skill, so your agent reads the play, asks before it spends anything, and hands the work back to you.

paste this into Claude
/hyreflow-recruit

Here is the spec for <ROLE_TITLE> at <CLIENT_COMPANY> in <LOCATION>.

<PASTE THE SPEC, OR GIVE A FILE PATH / URL>

Build me a shortlist of <N> candidates.

How I want it run:
- Search the general market and, separately, the direct competitors of
  <CLIENT_COMPANY>. Rank competitor-sourced people higher, but do not
  exclude anyone who is not at one.
- Pull each candidate's real work history before you score them. I do not
  want people ranked on their current job title alone.
- Score against the must-haves in the spec and drop anyone who fails them.
  Show me the reasoning.
- Only after that, find contact details for the ones who survived. Personal
  email and LinkedIn only. This is a candidate approach, so never work email.

Size the searches and tell me the cost before you spend anything. Deliver a
CSV plus a short summary of who is worth calling first and why.

Replace every <PLACEHOLDER> with your own detail. Everything else can stay as written.

What you need first

  • The job spec, as a file, a link, or pasted text
  • A Hyreflow workspace with credits, or your own provider keys connected
  • Optional: the client's competitor set, if you already know who to raid

Tools it can reach for

The agent picks per step from what your workspace has. Nothing here is required by name.

What happens when you run it

Free steps are marked free. Anything that spends credits is marked, and the agent asks before the first paid run of any size.

  1. 1

    Read the spec

    free

    Your agent parses the spec into a search: titles and their real-world variants, must-have skills, certifications, minimum experience, seniority band and the commute belt around the location.

  2. 2

    Work out who the competitors are

    credits

    Rather than guessing, the agent asks the web who actually competes with your client in that category, then resolves each to a domain worth searching inside.

  3. 3

    Search two ways at once

    credits

    One pass scoped to the competitor domains, one pass across the open market. The results merge and de-duplicate on LinkedIn profile, with competitor-sourced people ranked first.

  4. 4

    Pull the real work history

    credits

    A sourcing row carries a current title and employer and nothing else. The agent enriches the dated career history first, so tenure, trajectory and the actual scope of past roles are on the table before anyone gets scored.

  5. 5

    Score against the spec

    free

    Every candidate is judged against the must-haves with the evidence quoted back at you. Anyone who fails a must-have is dropped here, before a single contact-detail credit is spent.

  6. 6

    Find contact details for the survivors only

    credits

    Personal email and LinkedIn, walked through the enrichment providers cheapest first and stopping at the first hit. Addresses can be deliverability-checked before they reach a sequence.

  7. 7

    Hand it over

    free

    A CSV in your working folder, ranked, with the reasoning column intact. From here it can go into a sequence or straight into your ATS.

Why this beats a Boolean search and an export

A Boolean string filters on the words in a profile. It cannot tell you whether the person is any good, and it cannot tell you whether they are reachable. So the traditional shape of the job is: search, export a few hundred rows, buy credits to reveal contact details for all of them, then read through and discard most of what you paid for.

This play inverts the order. The expensive step, finding a personal email, runs last and runs only on people who already passed the spec. Qualification is the free part, so it runs on everyone. That single reordering is most of the cost difference between an automated desk and a manual one.

What you get back

A CSV in your own working folder with one row per candidate: name, current title and employer, the dated history that was used to score them, the tier or score, the reason, the personal email where one was found, and the LinkedIn profile. Alongside it, a short written summary naming who to call first.

Nothing is sent. Nothing is written into your ATS. The play stops at the shortlist and hands it to you.

Variations worth knowing

Name the competitors yourself. If you already know the four companies worth raiding, say so in the prompt. It saves a step and it is usually more accurate than anything the web will infer.

Ask for a bigger longlist. If the role is niche, ask for a wider top of funnel and a harder qualification threshold. You will spend a little more on history and considerably less on dead ends.

Add a competitor bench check. When four or more strong people surface at the same company, that is worth knowing on its own. Ask for the count by employer and you get a raid map alongside the shortlist.

Push it further. Once the shortlist looks right, the same session can draft the approach messages and enrol the candidates in a sequence. That is a separate, separately approved step, covered in build an outreach sequence.

Where this goes wrong

A vague spec produces a vague shortlist. The must-haves are what qualification scores against. If the spec says "strong communicator and a self-starter", there is nothing to score. Give it the two or three things that would actually make you reject a CV.

Very small employers have thin coverage. A five-person company leaves almost no trace in any people database. That is not a provider failing you can spend your way out of, so the play flags it rather than burning credits chasing it.

Do not filter on anything protected. Age, gender, ethnicity and the rest are not selection criteria and the play will not score on them, no matter how the prompt is worded. Score on skills, experience and evidence.

Questions

Why does it enrich the work history before scoring, when that costs money?

Because scoring on a job title is how you end up with a shortlist of people whose titles look right and whose experience does not. A row from a people database gives you a current title, a current employer and a location. The dated history tells you whether someone has actually done the thing for four years or was promoted into the title last month. Spending on history for a longlist is cheaper than spending on contact details for the wrong people.

Will it only give me people from the competitors?

No. The competitor pass is a ranking boost, not a filter. Running it as an exclusive filter is a good way to miss the strongest person in the market because they happen to work somewhere adjacent. Both passes run, both feed the same list, and the merge keeps the lineage so you can see where each name came from.

Why personal email and never work email for candidates?

You are approaching someone about leaving their current employer. Their work inbox is monitored, often filtered, and in a redundancy situation frequently already switched off. It is also the wrong etiquette, and candidates notice. Work email is the business-development channel, used when you are pitching a company, not a person.

How many candidates does it actually search to give me ten?

More than ten. Contact-detail coverage is a property of the person, not of how hard you try: a bare profile with no history and no public footprint returns nothing from any provider. So the play over-pulls at the top of the funnel, runs everyone through qualification, and delivers the best complete rows rather than padding the list with half-finished ones.

Can I run this against a spec I have not written yet?

Yes, if you can describe the role in a paragraph. The parse step is doing the same job either way. A real spec gives it better must-haves to score against, which is where the quality of the shortlist actually comes from.